Porosity and water permeability of rice husk ash-blended cement

composites reinforced with bamboo pulp

Conrado de Souza Rodrigues Khosrow Ghavami
Piet Stroeven
Received: 21 December 2004 / Accepted: 7 June 2005 / Published online: 16 September 2006
Springer Science+Business Media, LLC 2006
Abstract Cellulose bres have already been applied
commercially as an alternative to asbestos in bre-cements
composites. In spite of their industrial scale production for
more than 20 years, these composites still require much
research efforts, which focus mainly on durability aspects.
The inuence of the most relevant deterioration mecha-
nisms can be minimized if mineral admixtures with high
pozzolanic activity replace ordinary Portland cement
(OPC). The improvements then achieved are due to the
decrease in Ca(OH)
2
content and the more compact matrix
and interfaces in the composite. In this respect, rice husk
ash (RHA) is one of the most promising materials to be
applied as a partial cement replacement in the cellulose-
reinforced cement-based composites. This is due to the
high active silica content of the ash and the widespread
availability of the husks. To assess the inuences of dif-
ferent chemical compositions of RHA, and the effects of
autoclave curing on the pore characteristics of bamboo-
pulp-reinforced cement composites, a comparative study
was carried out in which pore characteristics were assessed
by mercury intrusion porosimetry (MIP). Complementa-
rily, the effects exerted by changes in the pore structure of
the composites on their water permeability are evaluated by
analytical and experimental approaches. It was observed
that the incorporation of RHA in the composites could
cause an extensive pore renement in the matrix and in the
interface layer, thereby decreasing water permeability. The
results indicate that partial replacement of cement by RHA
can improve the durability characteristics of cellulosece-
ment composites.
Introduction
Health problems associated with exposure to asbestos have
compelled countries since the early 70s to adopt policies
restricting the utilization of asbestos-based products. As a
consequence, asbestos-cement, one of the most successful
building materials, has been banned, or saw its employ-
ment restricted in an increasing number of countries. Due
to its high mechanical strength, lightweight, durability and
low cost, asbestos-cement components still play a key role
in low-cost housing projects, mainly in undeveloped and
developing countries. In trials to replace this material we
should therefore focus both on material behaviour and
affordability.
Among the wide range of available bres as a potential
replacement of asbestos, cellulosic pulps have always been
considered a viable alternative, when the physical and
chemical properties of the matrix, as well as the production
process of the composites are adjusted [1]. As a result,
cellulosic pulps have been applied on large scale as rein-
forcement of cementitious composites for more than two
C. de Souza Rodrigues (&)
Department of Civil Engineering, Federal Centre for
Technological Education of Minas Gerais - CEFET/MG,
Av. Amazonas, 7675, Belo Horizonte, CEP: 30510-000, Minas
Gerais, Brazil
e-mail: crodrigues@des.cefetmg.br
K. Ghavami
Department of Civil Engineering, Pontical Catholic University
of Rio de Janeiro, PUC-Rio, Av. Marques de Sao Vicente, 225,
CEP: 22453-900, Rio de Janeiro, RJ, Brazil
e-mail: ghavami@civ.puc-rio.br
P. Stroeven
Faculty of Civil Engineering and Geosciences, Delft University
of Technology, DUT, Stevinweg 1, 2628 CN, Delft,
The Netherlands
e-mail: p.stroeven@citg.tudelft.nl
J Mater Sci (2006) 41:69256937
DOI 10.1007/s10853-006-0217-2
1 3
decades [13]. These non-hazardous composites present
short-term mechanical and physical behaviour compatible
to that observed in asbestos-cement for most of the com-
mon applications. However, durability characteristics still
need considerable research efforts.
Environmental agents, combined with hydration prod-
ucts, activate a range of mechanisms in the high alkalinity
environment of the cementitious matrix that modify the
properties of the cellulosecement composites, impairing
their long term performance. These mechanisms are
fdescribed in detail in the literature [2, 48], being the
more direct effects on the composite strength due to:
Interfacial debonding: mainly due to bre-matrix abra-
sion that results from cycles of volumetric expansion-
shrinkage of the bres and matrix under temperature and
moisture variations.
Fibres embrittlement: resulting from migration of
hydration products to the bres pores and lumen.
Alkali attack: by some of the hydration products
causing dissolution of the cellulosic chain.
Since these deterioration mechanisms are strongly
dependent on moisture movement through the porous net-
work of the composites, one of the most applied preventive
methods is the partial replacement of cement by nely
ground mineral admixture with high pozzolanic activity.
The improved durability of the composites is due to the
lower matrix alkalinity and higher density of matrix and
interface layer [4, 7,8], resulting from the pozzolanic
reaction, involving Ca(OH)
2
consumption, and the lling
effect of the pozzolanic materials, respectively. Generally,
the addition of pozzolanic admixtures has chemical and
physical effects, although they are interrelated. The poz-
zolanic reaction can be understood in chemical terms as a
dissolution-precipitation mechanism, however yielding
reaction products smaller than the original admixture and
cement particles [9]. These act as ller, improving density
and enhancing physical (van der Waals-type of) binding
forces.
Rice husk ash (RHA) as a mineral admixture
Rice is a major crop in many world regions, accounting for
an estimated production of rice husk as an agricultural
residue totalling about 608 million tonnes in 2004 [10].
This amount would allow the generation of about
121 million tonnes of pozzolanic ash. Part of the husk is
applied as fuel during rice processing or for domestic
cooking in rural areas. However, most of this agricultural
by-product is simply disposed in nature representing an
environmental problem. Burning rice husks results in ash
with high silica content and, when the burning process is
carried out properly, the silica in RHA is amorphous,
making it a proper cement replacement material due to its
high pozzolanic activity [1113].
RHA is applied in this study as a partial replacement
material for OPC in cellulosecement composites as a
mean to obtain materials with improved durability char-
acteristics, which may be accomplished by composites with
denser and less permeable matrix and interfaces. Since the
RHA properties are very sensitive to the burning process,
ashes with different chemical compositions are considered
in this study. The composites were produced with different
replacement rates and subjected to conventional and
accelerated autoclave curing. To evaluate the effects of
these variables on the composite structure, a comparative
study was carried out applying MIP. The changes in the
pore structure due to RHA incorporation inuence the
transport mechanisms in the composite. Therefore, this
study also focused on water permeability of the compos-
ites, achieved analytically as well as experimentally.
Although only short-term porosity and permeability char-
acteristics are discussed, the modications in the material
structure may indicate improvements in their long-term
behaviour.
Porosity of cement-based materials
The porosity of a porous media, /, can be classied in
general terms in micro (d < 0.002 lm), meso
(0.002 < d < 0.05 lm) and macro pores (d > 0.05 lm)
[14, 15], where d is the pore diameter. In cement-based
materials, pores larger than 0.05 lm inuence strength and
permeability, while those smaller govern creep and
shrinkage behaviour [16]. In the mesoporosity range,
electrostatic interactions occur between the pore walls and
the uid, blocking transport through these pores. Hence,
the way mineral admixtures affect the transport properties
of cementitious materials can be interpreted in terms of the
increased volume of meso pores with a correspondent
decrease in that of macro pores [14].
Porosity is determined through MIP by measuring the
volume of mercury intruded in the sample under a given
pressure. During the intrusion test, the pressure is increased
from the atmospheric level to a couple of hundreds MPa,
resulting in intrusion curves reecting pressure data as a
function of the intruded volume of mercury. Under the
assumption that the pore structure consists of cylindrical
pores, the pressure applied to the mercury, P, is given in
terms of the pore diameter by the Washburn relation [17].
d
4:c:cosh
P
1
in which h is the contact angle between mercury and the
sample material, and c is the surface tension of mercury.
6926 J Mater Sci (2006) 41:69256937
1 3
MIP test data should be interpreted with great care,
accounting for the weak points of the method. Due to the
hypothesis of a cylindrical pores network and the uncer-
tainty of the contact angle adopted in Eq. (1), the pore
diameters calculated by the Washburn relation will not
represent the real values found in the sample [14, 18]. The
ink-bottle pores are represented in the intrusion curves
as several small pores, instead of a larger pore with a
narrow entrance. A third source of uncertainty in the
diameter values calculated by the Washburn relation is the
adoption of a single value as material property for the
contact angle, h, for composite samples with bended-
matrixes. Direct observation of the pore network of cement
pastes by scanning electron microscopy (SEM) resulted in
pore size distributions involving pore diameters 100
times larger than those from MIP tests [19], although this
approach also involves approximations, with three-dimen-
sional information being obtained from two-dimensional
images.
Despite all limitations, MIP is still employed as a
standard way to evaluate the pore structure of cement
materials [14] and regarded to be well reproducible [19,
20]. The method has shown a valuable tool in this com-
parative study to evaluate the effects of RHA incorporation
in the structure of the cellulosecement composites.
Permeability determination based on MIP tests
A laminar ow through a porous media can be expressed
by Darcys law,
q
K:DP
g:L
2
in which q is the ow rate (m
3
/s), DP is the pressure
gradient (N/m
2
), L is the length of the sample in the ow
direction (m), g is the uid viscosity (N s/m
2
), and K is the
permeability of the medium (m
2
) [21]. In spite of its sim-
plicity, the application of Eq. (2) for experimental deter-
mination of permeability is not an ordinary task, since the
laminar ow condition cannot easily be achieved in
experiments on low permeability materials. As a conse-
quence, the results usually reveal large discrepancies due to
dependence on the equipment and the experimental meth-
odology adopted [22].
The permeability of a medium is a function of its pore
structure. Models based on results from MIP tests have
quantitatively correlated both. One of these models was
developed by Katz and Thompson [23], here referred to as
the KT model and expressed by
K c:d
2
c
:
r
r
o
;
r
r
0

d
e
max
d
c
/:Sd
e
max
3
Herein, permeability is given in terms of the character-
istic diameter of the pore network, d
c
, and is r/r
0
the
relative conductivity, in which r is the electrical conduc-
tivity of a material saturated by a uid with electrical
conductivity r
0
. Finally, c is a constant of about 1/126 [23].
In the intrusion curves obtained from MIP tests, d
c
corre-
sponds to the threshold diameter, representing the opening
of the interconnected pore network in the sample.
The relative conductivity is related to the connectivity of
the pores in the sample and can also be established from
MIP data through the second equation in Eq. (3), whereby
d
max
e
is a characteristic dimension associated with maxi-
mum conductance. S(d
max
e
) is the porosity due to pores
larger than d
max
e
, and / is the total porosity. For materials
presenting a vast range of pore sizes, it is assumed that
d
max
e
= 0.34d
c
[21]. The KT model was originally applied
to sedimentary rocks, resulting in calculated permeability
data well correlated to values established experimentally
[23].
For cement-based materials, it was shown that the cal-
culated values of r/r
0
presents good correlation with
experimental values obtained through electrical conduc-
tivity measurements [24]. However, different attempts
reported in the literature have shown that the KT model
fails to predict the permeability of those materials, with
calculated values being in many cases two orders of
magnitude higher than the experimentally values [21,
2426]. Application of the KT model to cementitious
materials is criticized mainly because porosity is consisting
of capillary and gel pores, so that a single material
parameter, d
c
, cannot be used [27]. Therefore permeability
model on the basis of capillary and gel pores, by combining
these two phases through an effective media theory, was
developed by Cui and Cahyadi [26]. In this model, per-
meability is considered as a function of low and high
permeability phases, K
L
and K
H
, which are related to gel
and capillary phases, respectively, Eq. (4).
1 /
cap

K
1=t
L
K
1=t

K
1=t
L
AK
1=t

/
cap
K
1=t
H
K
1=t

K
1=t
H
AK
1=t
0 4
in which /
cap
is the porosity related to the high perme-
ability phase (pores larger than d
c
), t = 2 is assumed for
three-dimensional pore systems, and A = (1/
c
)//
c
, in
which /
c
is the critical capillary porosity value dened as
0.18 [28]. It is referred to in this study as the CC model.
The two permeabilities K
L
and K
H
follow from
K
H
c:d
2
c
1:8 1 /
c

2
; K
L
K
CSH
1
1 /
CSH
1 /
0
c
!
2
5
J Mater Sci (2006) 41:69256937 6927
1 3
The permeability of the capillary phase in the rst of Eq.
(5) is determined neglecting the permeability of the gel
phase, and is a function of d
c
only [26]. Next, K
L
is as-
sessed under the assumption that K
H
= 0 in Eq. (4), and
that the cement paste is composed of CSH, CH and
unhydrated particles, whereby CSH is the only perme-
able phase. Under such conditions Eq. (4) degenerates to
the second of Eq. (5), where K
CSH
and /
c
are the perme-
ability and the critical volume fraction of the CSH phase,
respectively, with approximate values of /
c
= 0.17 and
K
CSH
= 7.10
23
m
2
[26, 29].
The volume fraction of CSH, /
CSH
, is the ratio of the
volume of CSH, V
CSH
, and the total volume of the three
phases, i.e., in addition to CSH, CH, and the unhydrated
particles. The volumes of hydration products, V
CSH
+ V
CH
,
and the volume of unhydrated particles, V
uH
, are calculated
by means of [26, 29]:
V
CSH
V
CH

0:68a
w=c 0:32
;
V
uH

0:32 1 a
w=c 0:32
;
a
1 / :w=c 0:32/
0:36
6
Materials
RHA
A high-carbon-content RHA (RHA-I) was obtained from
Vietnamese rice husks burned for few seconds at temper-
atures ranging from 500 C to 750 C, as the ash pass
through a kiln specially developed for this purpose [12].
This high-carbon-content ash represents that resulted from
the application of the rice husk as a fuel. A low-carbon-
content RHA is produced (RHA-II) upon re-burning part of
this ash for 3 h in a conventional kiln at 750 C. There-
upon, both RHA samples were grounded in a ceramic ball
mill for 12 h.
Table 1 presents data on the chemical composition,
pozzolanic index DmS/cm, specic surface area obtained
by N
2
absorption, S
BET
, and the average particle size of the
ashes, d, determined through the Coulter Counter method
[30]. It can be inferred from Table 1 that a reduction of the
carbon content in the ash (and increasing the silica content)
causes signicant modications in the structure of the RHA
grains. RHA-I yields high S
BET
, which is not only due to
the smaller size of its particles, but is mainly the result of
the high-carbon-content particles of which the surfaces are
reported to be rich in micro porosities and discontinuities
[1113]. Also, it can be concluded that the re-burning
process resulted in a harder material, since RHA-II consists
of larger particles than RHA-I (Fig. 1), despite both being
subjected to the same grinding conditions. The silica
structure was also modied by the prolonged incineration
period applied to RHA-II, whereby a higher amount of
crystalline silica (crystoballite, C) was observed in the
X-Ray diffraction tests, as can be observed in Fig.1.
The pozzolanic activity of the ashes was determined by
measuring the change in the electrical conductivity of a
saturated lime solution as a result of RHA addition [31].
RHA-I adsorbs generally more Calcium ions from the lime
solution due to its higher S
BET
, resulting in a higher elec-
trical variation. So, it is as a consequence classied as a
highly pozzolanic material. On the other hand, RHA-II is
classied as moderately pozzolanic.
Bamboo-Pulp-reinforced cement composites
The composites were obtained by the slurry-dewatering-
compaction method simulating, in laboratory scale, the
industrial process for brecements production [32]. A
rened Kraft bamboo pulp with an average bre length of
Table 1 Chemical composition and physical properties of RHA and
OPC
RHA-I RHA-II OPC
SiO
2
(%) 73.6 94.6 17.9
C (LOI) (%) 22.9 1.4 1.6
Al
2
O
3
(%) 0.075 0.085 4.85
Fe
2
O
3
(%) 0.255 0.145 2.7
MgO (%) 0.27 0.280 2.15
CaO (%) 0.755 0.925 64.45
Na
2
O (%) 0.09 0.055 0.3
K
2
O (%) 1.165 1.115
SO
3
(%) 2.4
DmS/cm 1.4 0.58
S
BET
(m
2
/g) 119 15
d (lm) 1.8 4.0
Fig. 1 Particles size distributions and X-Ray diffraction results of
RHA-I and II
6928 J Mater Sci (2006) 41:69256937
1 3
1.47 mm was employed at a xed ration of 8% by dry mass
of the binder. The OPC employed was an ENCI cement
quality 32.5R (similar to an ASTM type I Portland ce-
ment), of which the composition is given in Table 1.
Partial cement replacements by RHA amounted 15%
and 30%; the different compositions of the specimens and
their mix codes being displayed in Table 2. The majority of
composites specimens were cured in a sealed plastic bag
for 7 days, and then exposed to the open air for the rest of
the curing period. This procedure is considered as the
standard curing method. For the composites subjected to
accelerated autoclave curing (8 h at 175 C, 1 MPa), an
A is added to the mix code. For comparison, plain ce-
ment matrices were similarly produced and, in this case,
only RHA-I was employed as a cement replacement
material.
Experimental
MIP tests
A Micromeritics PoreSizer 9320 performed the MIP tests
six months after the production of the composites and
pastes, with pressure varying from 0 MPa to 205 MPa. An
equilibrium period of 30 s was included in the procedure at
each pressure increment before recording the pressure and
intruded Hg volume data. Values of c = 485 dyn/cm
2
(0.1 Pa) and, based on commonly used values for contact
surface angle between Hg and cement, h = 141, were
adopted in Eq. (1). For comparative purposes, the intrusion
values are expressed in percentage of the bulk volume of
the sample.
Permeability of RHA-blended-cement pastes
and composites
The permeability of pastes and composites was determined
by KT and CC methods on the basis of the intrusion
curves obtained from MIP tests. Determination of the
volume fraction of CSH, /
CSH
, for assessment of K
L
in
the CC method, requires estimation of V
CSH
/V
CH
. For a
range of different cement pastes (w/c ratios of 0.3 and 0.4,
and hydration periods from 7 to 210 days) reported [26], a
xed V
CSH
/V
CH
value of 2.63 was applied. The same value
is used in this study for non-blended pastes and composites
(M0 and CP0). For the other composites, it is assumed that
the added RHA consumes all CH, so V
CH
= 0. The com-
posites were produced by the laboratory-scale Hatschek
method, without a xed w/c. However, a random check
after the slurry-dewatering-compaction process yielded the
w/c ratio to be about 0.3, which is the value applied in the
calculations for all blended composites.
Experimental evaluation of permeability
The permeability of some of the composites considered in
the porosity analysis was determined experimentally
applying a diffusion cell originally developed for the study
of transport phenomena in sedimentary rocks [33]. Disks
with a thickness of 6 mm thick and a diameter of 38.1 mm
were cut from the composite plates and placed between two
caps that were connected to the pressure pump and the
pressure transducers, so that the pressure on both upper and
lower surfaces of the sample could be continuously
recorded. The set-up of the sample in the diffusion cell is
depicted in Fig. 2a. The layers of porous stone between the
sample and the caps are for adjustment of the sample
thickness in the equipment, while the polyester cloth added
between the porous stone layer and cap at the top of the
sample compensates for small irregularities in the thickness
of the sample. This guarantees a perfect match between the
upper cap and the rest of the set-up. Unidirectional water
ow through the sample thickness was applied by the
pump. This condition was guaranteed by using a thermo-
retractable rubber layer around the whole set-up, perfectly
adjusted after submitting to heat. A metallic wire tightened
the two ends for extra security (Fig. 2b). The diffusion cell
was thereupon closed and lled with oil (Fig. 2c). This
allowed submitting the set-up to a constant conning
pressure of 3.0 MPa during the tests. Permeability can be
calculated by Darcys law after arriving at the laminar ow
condition under constant ow rate, Eq. (2). The main
inconvenience of this procedure is the long period required
to achieve the laminar ow through low permeability
materials. A method to determine permeability before the
laminar ow condition is established combines experi-
mental results and numerical modelling [33].
The upper surface of the sample is subjected during the
test to a constant water pressure, whereby the pressure
transducers record the way this pressure is transmitted to
the lower surface. In the numerical analysis, the sample is
Table 2 Composition and mix code of the samples
Mix code Binding (% dry mass) Pulp (% binding)
OPC RHA-I RHA-II
M0 100 0 0 0
M15 85 15 0 0
M30 70 30 0 0
CP0 100 0 0 8
CP15-I 85 15 0 8
CP30-I 70 30 0 8
CP15-II 85 0 15 8
CP30-II 70 0 30 8
CP30-IA 70 30 0 8
J Mater Sci (2006) 41:69256937 6929
1 3
approximated by a nite element network, whereupon the
numerical calculations are performed using the FPORO
program developed by the Petroleum Engineering and
Technology Group (GTEP/PUC-Rio) for permeability
studies of rocks on the basis of Darcys law. As input data,
the program requires information on the pressure imposed
experimentally, as well as on some physical properties of
the sample and of the applied uid, including an initial
value for permeability. The output of the numerical cal-
culation is a curve representing the variation of the pressure
on the lower surface of the sample due to the imposed
pressure on the opposite surface. An iterative procedure is
carried out whereby the input value of the permeability is
step-wise upgraded until the best correlation between the
numerical and experimental curves is achieved [34].
Results
MIP tests
Table 3 presents the physical properties of the composites
as derived from the MIP tests. Bulk density is the mass per
unit volume of the material with pores. The skeletal density
is the mass per unit volume of the pore-free material, i.e.,
ignoring the pores intruded by mercury. The maximum
cumulative intrusion volume is given in relation to sample
mass. The total porosity is the ratio of maximum intrusion
volume and bulk volume of the sample. Relying on the
reproducibility of the method [19, 20], only one sample
was tested per composition. Exceptions are plain paste
M15-I and composite CP15-II, of which two samples were
analyzed. Table 3 conrms these repeated tests to have
yielded close values for the properties calculated from
these tests. Also, the intrusion curves of these samples were
almost identical, supporting the reliability of the single-
sample set up.
Figure 3 presents the incremental and cumulative
intrusion curves of the plain cement matrix M0 and plain
composite CP0. The intrusion curves of M0 follow the
general trend indicated in the literature for cement pastes,
with no signicant intrusion being detected for pores larger
than about 0.2 lm and with an intrusion peak occurring at
this very level. This intrusion peak indicates the presence
of an inter-connected pore network and allows dening a
characteristic pore size, the threshold, or critical, pore
diameter [14, 16, 17]. The threshold pore diameter is re-
garded as an intrinsic characteristic of cement pastes, being
independent of the water/cement ratio [35]. Pores in the
0.05 lm to 0.15 lm range contribute by about 60% to
porosity of M0. A prominent peak at 0.12 lm implies 11%
of all pores in the sample to have a similar size. The pores
smaller than 0.05 lm account for about 21% of total
porosity.
The intrusion curves of CP0 reveal (Fig. 3) pores larger
than 0.15 lm responsible for the difference between the
pore structures of paste and composite. While in the paste
these pores contribute to about 8.5% of the total porosity,
in the composites they do so for almost 50%. The four
peaks occurring at pore diameters of 0.2, 0.25, 0.3 and
0.4 lm compose a pattern reproduced in all composites
tested in this study. Since pores smaller than 0.15 lm are
related to the matrix structure, and the main changes due to
RHA addition are observed in the range of pores smaller
than 0.4 lm, these peaks are probably coming from pores
in the interfacial transition zone, reecting a gradual pore
renement from the region closer to the bres to the pore
network of the bulk matrix.
The intrusion curves in Fig. 4 show the effects of the
addition of 30% RHA-I on the pore structure of the com-
posites. It can be observed that the blended-cement com-
Fig. 2 General set-up of the
permeability test: (a) the
composite sample is placed
between porous stones, (b) the
system is hindered laterally with
a thermo-retractable rubber
layer, and (c) the diffusion cell
is closed lled with oil,
Table 3 Physical properties resulted from MIP tests
Sample
composition
Bulk
density
(kg/m
3
)
Skeletal
density
(kg/m
3
)
Max. Cum.
Int. Vol.
(m
3
/kg) (10
3
)
Porosity
(%)
M0 1972.1 2297.4 0.072 14.2
M15-I (a) 1778.1 2148.7 0.097 17.2
M15-I (b) 1779.9 2135.2 0.094 16.6
M15-I (mean) 1779.0 2142.0 0.095 16.9
M30-I 1628.3 1951.5 0.102 16.6
CP0 1613.1 2146.3 0.154 24.8
CP15-I 1505.0 2013.5 0.168 25.3
CP30-I 1361.5 1829.6 0.188 25.6
CP15-II (a) 1504.6 1968.0 0.157 23.5
CP15-II (b) 1479.7 1930.4 0.159 23.3
CP15-II (mean) 1492.2 1949.2 0.157 23.4
CP30-II 1503.2 1912.2 0.142 21.4
CP30-IA 1376.0 1834.7 0.182 25.0
6930 J Mater Sci (2006) 41:69256937
1 3
posites have relatively high volume of pores larger than
0.15 lm, distributed in a similar pattern as that observed in
the plain composite. The main effects of RHA-I can be
observed on pores smaller than 0.15 lm, which are char-
acteristic of the matrix. The blended-cement composites
show lower volume of pores with diameter between 0.05
and 0.15 lm. The reduced porosity of the matrix is prob-
ably an effect of the pozzolanic reaction, with its products
blocking the pore network. Additionally, the intrusion
curves from samples of composites with 15% and 30%
RHA presented similar features; therefore, only the intru-
sion curves from composites with 30% RHA are shown.
In relation to CP0, the addition of RHA-II decreases
both density and porosity of the composites (Table 3). It is
revealed in Fig. 5 that 30% RHA-II decreases the porosity
related to the localized peaks (interface) and, mainly, that
related to the matrix. Instead of the intrusion peak that
characterizes the threshold diameter of the matrix in the
plain-cement composite, the blended-cement composites
present intrusion levels remarkably lower. Again, once the
RHA particles are larger than the pores in the rened range
(0.080.3 lm) this renement may be the effect of porosity
blocking by the products of the pozzolanic reaction, which
seems to be enhanced by the decreased carbon content in
RHA-II.
Although the small changes in density and porosity
values in Table 3 suggest little inuence of the curing
process on the pore characteristics of the composites, the
intrusion curves in Fig. 6 show the opposite. The effects of
autoclave curing on pore characteristics of composites with
RHA-I are similar to those resulting from addition of RHA-
II in composites under standard curing, as presented in
Fig. 5. A renement of porosity in the interfacial region,
and a strong modication in the size distribution of the
pores of the matrix are observed in Fig. 6. Pore sizes
around the threshold level are responsible for only a
modest intrusion volume, but there is an increase in the
intrusion rates at ner pore levels, showing the renement
of the matrix due to the autoclave process.
Calculated permeability of cement pastes and
composites
A large discrepancy between the results obtained from the
CC and KT models, can be seen in Table 4. Comparing
the obtained results in Fig. 7, with those from the available
Fig. 3 Comparative intrusion
curves of matrix (M0) and
composite (CP0)
Fig. 4 Effects of RHA-I
incorporation on the pores
distribution of composites
J Mater Sci (2006) 41:69256937 6931
1 3
published data, indicates that the CC model represents the
matrix permeability more realistically. The KT model
results in permeability, about 10
19
m
2
, that is, about 10
times higher than data reported for pastes with w/c of 0.25
and 0.36 and hydrated for 182 days, 0.4 10
20
m
2
and
10
20
m
2
, respectively [21]. The latter value is in the same
order of magnitude as the permeability data for 28-days-old
cement pastes with w/c of 0.47 [25]. However, after
210 days, cement pastes with w/c of 0.3 and 0.4 presented
permeabilities of 2.8 10
23
and 1.3 10
22
m
2
, respec-
tively [26]. Furthermore, it should be noted that total
porosities of the cement pastes reported are lower than the
critical value of 18%, dened by Garboczi and Bentz [28]
as the value below which connectivity of the pores
decreases sharply. Based on the presented data it can be
assumed that the CC model better represents the perme-
ability of the cement pastes. It could be concluded that the
addition of RHA-I does not modify signicantly the per-
meability of the cement pastes, in the order of 10
22
m
2
.
For the studied cement pastes, the threshold diameter,
d
c
, is easily identied, as far as almost no signicant
intrusion is observed in pores larger than d
c
. The intrusion
curves of the bamboo-pulp-reinforced composites are
more complex, reecting features characterizing the
matrix, the interface and the reinforcement. The com-
plexity of the pore structure makes it difcult to dene a
single parameter that would represent realistically the
whole pore network. As a rst approximation, the threshold
diameter of the composite can be assumed similar as that
of the matrix (about 0.1 lm), whereas the additional
porosity resulting from the incorporation of the reinforce-
ment will consist of larger pores in the same network.
This rendered possible calculating by KT and CC
models the permeability data listed in Table 4 and plotted
in Fig. 7.
Data regarding a commercial cellulosecement com-
posite reported in the literature [7] refers to permeability to
Nitrogen at the level of 10
18
m
2
. The KT model results
values of the same order of magnitude. However, the
results are almost the same for all composites considered,
showing that the renements in the pore structure observed
in some composites do not inuence permeability. The
CC model is more sensible to variations in pore structure
of the composites. However, the measured permeability of
the composites can be considered too low when compared
to permeability for cement pastes reported in the literature
[21, 25, 26] and to the aforementioned information
presented for cellulosecement composites [7].
Fig. 5 Effects of RHA-II
incorporation on the pores
distribution of composites
Fig. 6 Effects of accelerated
autoclave curing on the pores
distribution of composites with
RHA-I
6932 J Mater Sci (2006) 41:69256937
1 3
Analysing the intrusion curves of the composites,
Figs. 36, and the models for permeabilty, it can be
observed that when larger pores are regarded as d
c
, the
discrepancies between the results of the two analytical
models increase. The scale of the threshold diameter in Eq.
(3) denes the magnitude of the results of the KT model.
As a consequence, larger threshold diameters will result in
higher permeability. On the other hand, the magnitude of
the permeability in the CC model depends mainly on the
porosity adopted in the effective media equation. In Eq. (5)
the porosity is a weight function reecting the relative
importance of the low and high permeability phases. In the
cumulative intrusion curves, larger threshold diameters
correspond to lower porosity, which in turn enhances the
contribution of the low permeability phase in Eq. (5),
resulting in the lower values of permeability. The analyti-
cal models therefore fail to predict the permeability of the
composites, in the cases studied.
Modied model for permeability of composites
To obtain more realistic results for the permeability of the
composites, a modied analytical model is proposed. It is
also based on the effective media theory, considering the
properties of the intrusion curves of the composites. It
considers porosity due to pores larger than 0.2 lm as
related to the interface zones and to the reinforcement,
being the high permeability phase, K
H
*
, and the porosity
corresponding to the matrix being the low permeability
phase, K
L
*
, as shown in Fig. 8. This modication of the Cui-
Cahyadi [26] model is referred to in this study as the CC
*
model.
The permeability K
H
*
is calculated by means of the KT
model, Eq. (3), with pores of about 5 lm being regarded as
d
c
, whereby this pore size represents the rst deection of
the cumulative intrusion curves of all composites, i.e.
interconnected larger pores. For the determination of K
L
*
,
the CC model is applied, Eqs. (4) to (6), in which the pore
structure is considered consisting of pores smaller than
0.2 lm, and with a threshold diameter equal to that of the
matrix. Once K
H
*
and K
L
*
are established, the permeability
of the composite is calculated by Eq. (4). By denition, the
porosity, /, in Eq. (4) is that of the high permeability
phase. This sub-division in the pore structure in the CC*
model indicates that / should be composed of pores larger
than 0.2 lm. However, it is stated that changes in the
permeability of cement pastes due to blending occur as a
result of the renement of their pore structure, i.e., by
Table 4 Calculated permeability of pastes and composites through analytical models
d
c
(lm) / KT model CC model
d
max
e
(lm)
S(d
max
e
) (r/r
0
) K (m
2
)
10
2
/
c
a V
CSH
V
nH
/
CSH
K
H
(m
2
)
10
16
K
L
(m
2
)
10
23
K (m
2
)
10
2
M0 0.11 0.14 0.04 0.11 0.0055 29.225 0.06 0.73 0.84 0.12 0.87 0.79 4.14 0.009
M15-I 0.10 0.17 0.04 0.12 0.0069 32.126 0.04 0.76 0.84 0.12 0.87 0.67 5.00 0.009
M30-I 0.08 0.17 0.03 0.12 0.0065 19.832 0.06 0.74 0.81 0.14 0.86 0.45 4.79 0.009
CP0 0.13 0.25 0.043 0.21 0.0177 125.263 0.15 0.57 0.45 0.22 0.67 1.04 2.54 0.081
CP15-I 0.13 0.25 0.043 0.21 0.0182 128.297 0.17 0.54 0.60 0.23 0.72 1.04 3.05 0.446
CP30-I 0.13 0.26 0.043 0.20 0.0174 122.436 0.17 0.54 0.60 0.23 0.72 1.04 3.05 0.446
CP15-II 0.13 0.23 0.043 0.19 0.0155 109.002 0.16 0.56 0.61 0.23 0.73 1.04 3.15 0.192
CP30-II 0.13 0.21 0.043 0.17 0.0120 84.871 0.14 0.60 0.65 0.21 0.76 1.04 3.51 0.066
CP30-IA 0.13 0.25 0.043 0.16 0.0134 94.253 0.13 0.61 0.66 0.20 0.77 1.04 3.60 0.045
Fig. 7 Comparative values for
the permeabilities of matrix and
composites
J Mater Sci (2006) 41:69256937 6933
1 3
decreasing the volume of the macro pores and by
increasing the volume of meso pores [14]. When this is
taken into account, the border between macro and meso
pores can be considered the minimum pore diameter of
capillary porosity, whereby / is associated with pores
larger than 0.05 lm.
The results from the CC* model are shown in Table 5
and in Fig. 7. It can be observed that differences with the
data reported in the literature [7] are reduced. The model is
also sensitive to the renement of the pore structure
resulting from the addition of RHA-II, and to the autoclave
curing. These modications lead to 1000 times lower
permeability data of the order of about 10
21
m
2
.
Experimental determination of permeability
By applying a constant pressure on the upper surface of the
sample (P. Top), and observing the different rates at which
this pressure is transmitted to the lower surface (P. Bot-
tom), it is possible to evaluate qualitatively how the dif-
ferent pore structures inuence the transport properties of
the composites. Fig. 9ad presents information on pressure
variations with time on both upper and lower surfaces of
the composites for three different consecutive pressure
levels. It can be seen that the pressure transmission from
the upper surface to the opposite one occurred in a similar
manner in normal-cured composites with and without
RHA-I, CP0 and CP30-I, respectively, Fig. 9a, b. A lower
transmission rate is observed in the autoclaved composite,
CP30-IA, Fig. 9c, showing that this composite presents a
signicantly lower permeability. For the blended compos-
ites containing 30% low-carbon-content RHA, CP30-II,
Fig. 9d, it was observed that the pressure is transmitted
through the sample at a much lower rate than in the other
composites, demonstrating the transport properties of the
composite to be strongly inuenced by the addition of this
type of RHA.
Permeability of the composites was assessed by means
of the FPORO program, based on the experimental curves
obtained in the diffusion cell. In Fig. 9ad it can be seen
that the curves resulting from the numerical calculations
are well correlated to those obtained experimentally. The
permeability values that generate these numerical curves
are presented in Table 6 and Fig. 7, demonstrating the
addition of 30% RHA-I slightly to increase permeability of
the composites.
MIP tests revealed that autoclaved composites with high-
carbon-content RHA and composites with low-carbon-
content RHA submitted to standard curing have pore
structures with similar characteristics, with a rened
porosity in the phases related to the interface zones and to
the matrix, Figs. 5 and 6. Based on such pore characteristics,
Fig. 8 Low and high
permeability phases of cement
composites considered in the
modied analytical model
Table 5 Permeabilities calculated by the modied analytical model
K
H
*
: KT model K
L
*
: CC model CC* model
d
c
(lm)
/ d
max
e
(lm)
S(d
max
e
) (r/r
0
) K
H
*
(m
2
)
10
16
d
c
(lm)
/
c
a V
CSH
V
nH
/
CSH
K
H
(m
2
)
10
16
K
L
(m
2
)
10
23
K
L
*
(m
2
)
10
23
/* K* (m
2
)
10
2
CP0 4.00 0.13 1.36 0.08 0.004 2.66 0.13 0.04 0.76 0.84 0.12 0.87 1.04 5.03 8.30 0.21 43.000
CP15-I 4.00 0.15 1.36 0.09 0.005 3.39 0.13 0.04 0.77 0.84 0.12 0.88 1.04 5.06 8.30 0.21 53.800
CP30-I 4.00 0.15 1.36 0.09 0.005 3.40 0.13 0.04 0.76 0.84 0.12 0.87 1.04 5.01 8.30 0.20 28.000
CP15-II 3.50 0.16 1.19 0.11 0.006 3.29 0.13 0.02 0.80 0.88 0.10 0.90 1.04 5.36 6.80 0.19 10.900
CP30-II 3.00 0.13 1.02 0.09 0.004 1.55 0.13 0.02 0.79 0.87 0.11 0.89 1.04 5.26 6.70 0.16 0.390
CP30-IA 4.00 0.13 1.36 0.08 0.003 2.38 0.13 0.02 0.80 0.88 0.10 0.89 1.04 5.33 6.70 0.15 0.083
6934 J Mater Sci (2006) 41:69256937
1 3
it could be expected that composites CP30-IA and CP30-II
would yield according to the modied analytical model
close values for permeability, i.e., lower than that of the
other samples. However, it is observed that permeability of
composite CP30-II is much lower than that of the autoclaved
composite CP30-IA. This obvious mismatch could be due to
the roughness of the pore surfaces resulting from different
curing conditions. Moreover, porosity measurements were
performed under quasi-static conditions, while permeability
tests were conducted in a situation of dynamic water ow.
As a result, crystallization of hydration products at auto-
clave curing may lead to smoother pores surfaces (lower
tortuosity), offering a lower resistance to water movement
during the permeability tests. On the other hand, the normal
cured CP30-II may feature more irregularly shaped pores
surfaces (higher tortuosity), reected by reduced perme-
ability.
Concluding remarks
This comparative study offers supporting evidence for
using MIP, in spite of all limitations, as a valuable tool
in assessing modifying effects of mineral admixtures
and curing conditions on pore characteristics of cement
composites. The main conclusions as to these effects are
the following:
The different replacement rates of OPC by RHA
applied (15% and 30% RHA) exert little inuence on
the pore structure of the composites as reected by
comparable intrusion curves.
Density is reduced and porosity increased when high-
carbon-content RHA (RHA-I) is incorporated in the
composites. Total porosity increases probably as a
consequence of high porosity in the surface layer of the
ash particles. Cement composites blended with this ash
yield higher volume of pores exceeding 0.15 lm and
lower volume of pores at the threshold diameter of the
matrix, demonstrating the modest renement in the
pore structure of the matrix.
Addition of low-carbon RHA (RHA-II) strongly mod-
ies the pore size distribution in the interface zones and
in the matrix, leading to an overall renement. The
decreased porosity probably results from the enhanced
chemical interaction between the low-carbon-content
RHA and the cement in the matrix.
Results of the different RHA on the pore structure of
cellulosecement composites seem related to their dif-
ferent pozzolanic activities, since the more signicant
differences are observed in the pore range smaller than
Fig. 9 Pressure transmission curves resulted from the permeability tests
J Mater Sci (2006) 41:69256937 6935
1 3
the RHA particles. As a consequence, the higher poz-
zolanic activity (of RHA-II) results in a higher amount
of reaction products that decreases the porosity by
deposition. This opposes to the results in Table 1,
which presents RHA-I as more reactive. Due to its
higher S
BET
value, and the consequent adsorption of
more Calcium ions, RHA-I results in more variation in
the electrical conductivity of the saturated lime solution
than RHA-II. However, it does not mean that all the
adsorbed Calcium reacts with the silica in the ash. The
more pronounced effects of RHA-II on the pore net-
work and permeability of the composites, despite its
larger particles, lower S
BET
and higher crystalline silica
content, shows that higher carbon content may prevent
a proper pozzolanic reaction between the mineral
admixture and cement reaction products.
Blending with high-carbon-content RHA can yield
similar effects as observed with RHA-II-blended com-
posites provided they are subjected to accelerated
autoclave curing.
Based on general features of the intrusion curves
observed for all composites, it is proposed to sub-divide
pore structure in three phases: pores with diameters
larger than 0.5 lm, introduced by the reinforcement;
pores in the 0.2 and 0.5 lm diameter range, which
would represent the interfacial zones, and pores smaller
than 0.2 lm, characterizing the porosity of the matrix.
Regarding the relation between the pore network and the
permeability of the cement composites, the analytical
approaches presented in the literature, originally developed
for permeability prediction of rocks and cement pastes,
were not properly correlated to the features reected by the
intrusion curves of the composites. A modication on these
models resulted in more realistic estimates for the perme-
ability, agreeing with the experimental results of normal-
cured composites. However, for autoclaved composites, the
experimental permeability does not correspond to the value
obtained from the analytical model. This is probably due to
different pore surface characteristics resulted from the
autoclaved curing, with the composites showing a similar
pores distribution to that of some normal cured composites
(during the quasi-static MIP test) and quite different per-
meability behaviour (during the dynamic diffusion test).
Finally, the porosity and permeability results allow a
better insight into mechanisms to improve by RHA incor-
poration the durability of composites. The porosity of
matrix and interface is decreased, reducing water perme-
ability, probably preventing or minimizing the effects of
the main transport-related deterioration mechanisms
affecting these composites under normal service life con-
ditions. In this sense, best performance is expected from
normal cured composites blended with low-carbon RHA.
Acknowledgements This study was nancially supported by
CNPqNational Council for Scientic and Technology Development,
Brazilian Ministry of Science and Technology, and was developed at
Delft University of Technology as a cooperation project supported by
CICATManagement Centre for International Cooperation. The au-
thors are grateful to Doctor Eudes Siqueira Muniz, from the Petro-
leum Engineering and Technology Group-GTEP/PUC-Rio, for
making possible the experimental study on permeability with the
diffusion cell.
References
1. Studinka JB (1989) Int J Cem Lightweight Concr 11:73
2. Coutts RSP (1988) In: Swamy RN (ed) Natural bre reinforced
cement and concrete. Blackie and Son Ltd, Glasgow, 1pp
3. Sharman WR, Vautier BP (1986) Dur Build Mat 3:255
4. Gram HS, Persson H, Skarendhal A (1984) In: Natural Fibre
Concrete (SAREC, Stockholm) 5pp
5. Soroushian P, Marikunte S (1984) In: Daniel JI, Shah SP (eds)
Fiber reinforced concrete: developments and innovations.
American Concrete Institute, 73pp
6. Canovas ME, Selva NH, Kawiche GM (1992) Mat Struct 25:417
7. Macvicar R, Matuana LM, Balatinecz JJ (1999) Cem Concr
Comp 21:189
8. Toledo Filho RD, Ghavami K, England GL, Scrivener K (2003)
Cem Concr Comp 25:185
9. Yu Q, Sawayama K, Sugita S, Shoya M, Isojima Y (1999) Cem
Concr Res 29:37
10. FAO (2004) In: Rice market monitor, vol VII, issue 4, (Food and
Agriculture Organization for the UN, http://www.fao.org/es/ESC/
en/20953/21026/highlight_23001en_p.html, 9th Sept)
11. Cook DJ (1986) In: Swamy RN (ed) Cement replacement mate-
rials. Blackie & Son Ltd, London, 171pp
12. Stroeven P, Bui DD, Sabuni E (1999) Fuel 78:153
13. Bui DD (2001) In: Rice husk ash as a mineral admixture for high
performance concrete, Delft University Press, Delft, 13pp
14. Brown PH, Shy D, Skalny J (1991) In: Skalny J, Mindess S (eds)
Materials science of concrete II. The American Ceramic Soc. Inc,
83pp
15. Aldea CM, Young F, Wang K, Shah SP (2000) Cem Concr Res
30:465
16. Metha PK, Monteiro JP (1993) In: Concrete: structure, properties
and methods. Prentice Hall, New Jersey, 26pp
17. Winslow DN, Lovell CW (1981) Powder Tech 29:151
Table 6 Permeabilities established through the coupled experimental-
numerical solution
Sample P Top (kPa) K (m
2
)
10
2
K
mean
(m
2
)
10
2
CP0 1800 18.50 21.0
1500 18.50
1000 26.0
CP30-I 1800 35.0 38.0
1500 40.0
1000 4.00
CP30-IA 1800 4.50 6.17
1500 6.50
1000 7.50
CP30-II 1800 0.45 0.38
1500 0.30
1000 0.40
6936 J Mater Sci (2006) 41:69256937
1 3
18. Diamond S (2000) Cem Concr Res 30:1517
19. Ye G (2003) In: Experimental study & numerical simulation of
the development of the microstructure and permeability of
cementitious materials. Delft University Press, Delft, 43pp
20. Hedenblad G (1997) Adv Cem Based Mat 6:123
21. El-Dieb AS, Hooton RD (1994) Cem Concr Res 24:443
22. Banthia N, Mindess S (1989) Cem Concr Res 19:727
23. Katz AJ, Thompson AH (1986) Phys Rev B 34:8179
24. Tumidajski PJ, Lin B (1998) Cem Concr Res 28:643
25. Christensen BJ, Mason TO, Jennings HM (1996) Cem Concr Res
26:1325
26. Cui L, Cahyadi JH (2001) Cem Concr Res 31:277
27. Garboczi EJ (1990) Cem Concr Res 20:591
28. Garboczi EJ, Bentz DP (1996) Constr Build Mat 10:293
29. Bentz DP, Garboczi EJ (1991) Cem Concr Res 21:325
30. Allen T, Marshall K (1972) In: The electrical sensing zone
method of particle size measurement: the Coulter principle.
University of Bradford, England, 105pp
31. Luxan MP, Madruga F, Saavedra J (1989) Cem Concr Res 19:63
32. Campbell MD, Coutts RSP (1980) J Mat Sci 15:1962
33. Muniz ES (2003) In: Desenvolvimento de Equipamento e Me-
todologia de Testes para Avaliacao da Interacao Folhelho-Fluido
de Perfuracao. Doctorate thesis, Pontif cia Universidade Catolica
do Rio de Janeiro, Rio de Janeiro, in Portuguese, 41pp
34. Frydman M, Fontoura SAB (2001) In: Proceedings of the Latin
American and Caribbean Petroleum Engineering Conference.
Buenos Ayres, 8pp
35. Matte V, Moranville M (1999) Cem Concr Comp 21:1
J Mater Sci (2006) 41:69256937 6937
1 3